A race condition in the Linux kernel's swap cache can cause data corruption when two or more threads swap in the same entry at the same time. This can lead to a stalled page being installed into the page table, resulting in data loss. The issue is triggered when multiple threads swap in the same entry simultaneously, causing the `pte same` check to pass even though the page has been modified. This can happen when using a small swap device and multiple threads are updating mapped pages in opposite directions. A reproducer has been created to demonstrate this issue, which can cause data loss at a rate of about once every 5 minutes.